antv-s2-expert

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkDeclared5 observations
• ShellNone observedNo commands run in the source
• DependenciesNot pinnedSome versions float
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

const header = `<div style="background:#5B8FF9;color:#fff;padding:6px 8px;font-weight:bold">${tableName || 'table'}</div>`;

const header = `<div style="background:#1890ff;color:#fff;padding:6px 8px;font-weight:bold;">${data.name || 'Table'}</div>`;

const header = '<div style="background:#1890ff;color:#fff;padding:6px 8px;font-weight:bold;">' + (data.name || 'Entity') + '</div>';

const header = '<div style="background:#1890ff;color:#fff;padding:6px 8px;font-weight:bold;">' + (data.name || 'Entity') + '</div>';

dotenv, esbuild, pngjs, puppeteer

@antv/context, commander, gray-matter, @types/node, tsx, typescript, vitest

eval/cases/g2-dataset-174.json

Audit 0.4.1 · source sha 34615c1ed06b · 22 Oct 2026Report an issue or request a re-scan

• Publisherantvis GitHub organizationOwner of the repository this skill is read from
• Source folderantvis/chart-visualization-skillsThe exact folder we audited
• LicenceMITAs the repository states it
• Commit pinned for this verdict34615c1ed06bA new audit runs when the source changes
• Repository statusNot archivedLast push 30 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Use this skill whenever the user wants to create, customize, or troubleshoot G2 v5 chart visualizations. Triggers include: any mention of 'G2', 'antv g2', '@antv/g2', 'G2 chart', 'G2 可视化', or requests to produce charts…

---
name: antv-g2-chart
description: "Use this skill whenever the user wants to create, customize, or troubleshoot G2 v5 chart visualizations. Triggers include: any mention of 'G2', 'antv g2', '@antv/g2', 'G2 chart', 'G2 可视化', or requests to produce charts like bar charts (柱状图), line charts (折线图), pie charts (饼图), scatter plots (散点图), area charts (面积图), heatmap (热力图), radar charts (雷达图), treemap (矩形树图), funnel charts (漏斗图), sankey diagrams (桑基图), gauge (仪表盘), wordcloud (词云), boxplot (箱线图), as well as G2-specific topics like encode channels, scale config, coordinate systems, transforms, interactions, themes, labels, and animations. Also use when debugging G2 rendering errors, V4→V5 migration issues, or chart type selection. Do NOT use for G6 graph/network visualization, X6 editor diagrams, or S2 pivot tables."
tools:
  - curl
---

# G2 v5 Chart Visualization

## Overview

G2 v5 is AntV's grammar-of-graphics charting library. It uses **Spec Mode** — a declarative, JSON-like configuration style. Generated examples should start from a complete spec; runtime code may later merge a local update through `chart.options()`.

```javascript
import { Chart } from '@antv/g2';

const chart = new Chart({ container: 'container', autoFit: true });
chart.options({
  type: 'interval',
  data: [{ genre: 'Sports', sold: 275 }],
  encode: { x: 'genre', y: 'sold' },
});
chart.render();
```

### CDN Usage

```html
<script src="https://unpkg.com/@antv/g2@5/dist/g2.min.js"></script>
<script>
  const chart = new G2.Chart({ container: 'container', autoFit: true });
  chart.options({
    type: 'interval',
    data: [{ genre: 'Sports', sold: 275 }],
    encode: { x: 'genre', y: 'sold' },
  });
  chart.render();
</script>
```

## Content Retrieval Service

When using AntV G2 for data visualization, if you need to understand the concepts, usage, API, examples, and other aspects of G2 v5, you can use the provided context retrieval service. When using the skill, content is retrieved via an antv HTTP API server using GET requests.

- Host: `https://sive.antv.antgroup.com`
- Endpoint: `/api/v1/context/retrieve`
- Method: `GET`
- Parameters: `query`, `library`, `topK`, `content`, `maxTokens`

Retrieve reference documents by query (hybrid search = FTS + vector + RRF fusion).

| Parameter | Type | Required | Description |
|---|---|---|---|
| `query` | string | ✅ | Search keywords, e.g. `bar chart interval` |
| `library` | string | ✅ | Library name: `g2`, `g6`, `x6` |
| `topK` | number | | Number of results to return (default: 5) |
| `content` | boolean | | Return full reference doc markdown (default: true) |
| `maxTokens` | number | | Max tokens per result (default: unlimited) |

```bash
curl "https://sive.antv.antgroup.com/api/v1/context/retrieve?query=bar+chart+stacked&library=g2"
```

## Critical Rules

### MUST: Use V5 Spec Mode ONLY

```javascript
// ❌ WRONG — V4 chain API (deprecated, will not render)
chart.interval()
  .data([...])
  .encode('x', 'genre')
  .encode('y', 'sold')
  .style({ radius: 4 });

// ✅ CORRECT — V5 Spec Mode
chart.options({
  type: 'interval',
  data: [...],
  encode: { x: 'genre', y: 'sold' },
  style: { radius: 4 },
});
```

### Multi-mark overlays use `view + children`

`chart.options()` supports incremental deep-merge updates, which is useful for runtime interaction. For generated one-shot code, provide a complete initial spec. To create independent overlay marks, use one `type: 'view'` with `children`; sequentially changing the root `type` does not create an overlay.

```javascript
// ❌ WRONG — this changes one root mark from line to point; it does not overlay them
chart.options({ type: 'line', data, encode: { x: 'date', y: 'value' } });
chart.options({ type: 'point', data, encode: { x: 'date', y: 'value' } });

// ✅ CORRECT — children array for multi-mark
chart.options({
  type: 'view',
  data,
  children: [
    { type: 'line',  encode: { x: 'date', y: 'value' } },
    { type: 'point', encode: { x: 'date', y: 'value' } },
  ],
});
```

### MUST: `container` is mandatory, `chart.render()` at the end

```javascript
// ❌ WRONG — no container, no render
const chart = new Chart();
chart.options({ type: 'interval', data });

// ✅ CORRECT
const chart = new Chart({ container: 'container', autoFit: true });
chart.options({ type: 'interval', data, encode: { x: 'genre', y: 'sold' } });
chart.render();
```

### MUST: Correct mark types only

| ❌ Hallucinated (from ECharts/Vega) | ✅ G2 correct replacement |
|---|---|
| `type: 'ruleX'` | `type: 'lineX'` |
| `type: 'ruleY'` | `type: 'lineY'` |
| `type: 'regionX'` | `type: 'rangeX'` |
| `type: 'regionY'` | `type: 'rangeY'` |
| `type: 'venn'` | `type: 'path'` + transform |

**Legal G2 marks**: `interval`, `line`, `area`, `point`, `rect`, `cell`, `text`, `image`, `path`, `polygon`, `shape`, `link`, `connector`, `vector`, `lineX`, `lineY`, `rangeX`, `rangeY`, `range`, `box`, `boxplot`, `density`, `heatmap`, `beeswarm`, `treemap`, `pack`, `partition`, `tree`, `sankey`, `chord`, `wordCloud`, `gauge`, `liquid`. `sunburst` requires `@antv/g2-extension-plot`.

### MUST: `encode` is an object, `transform` is an array

```javascript
// ❌ WRONG
.encode('x', 'genre')
.transform: { type: 'stackY' }

// ✅ CORRECT
encode: { x: 'genre', y: 'sold' }
transform: [{ type: 'stackY' }]
```

### MUST: `labels` is plural; range encodings use mark-appropriate channels

```javascript
// ❌ WRONG
label: { text: 'sold' }

// ✅ CORRECT
labels: [{ text: 'sold' }]
// Interval / link ranges may use either a two-field array or y + y1.
encode: { y: ['start', 'end'] }
// RangeY explicitly uses y + y1.
encode: { y: 'start', y1: 'end' }
```

### MUST: No d3 in user code

```javascript
// ❌ WRONG — d3 is not exposed in user scope
const total = d3.sum(data, d => d.value);

// ✅ CORRECT — use native JS or G2 built-in transforms
const total = data.reduce((sum, d) => sum + d.value, 0);
```

### MUST: No white/near-white fill, no `padding` as array

```javascript
// ❌ WRONG
style: { fill: '#fff' }  

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

1. 2026-10-22
   !CAUTION B 89 / 100

   first audit · source 34615c1ed06b

Questions about antv-s2-expert

• github.com/antvis/chart-visualization-skills — the audited source at commit 34615c1ed06b and every finding aboveObserved · 22 Oct 2026

Spotted something wrong? Report a correction · Methodology